Thanks, folks! Everything you've said has clarified many things, and I'm
working on putting them into practice. But in the meantime, I've got a new
error condition, with the same repositories (and complex repository
dependencies), and could use some guidance.
I started to see this error condition when I removed the toolshed and
changeset_revision tags in the tool_dependencies.xml for the dependent
repository xena_install. It now reads:
<package name="installXena" version="1.0">
<repository name="start_xena" owner="melissacline"/>
Short and simple. But lately, when I try to install this repository from
the test toolshed, I get a pop-up window from Chrome (my web browser) that
says "The page at 127.0.0.1:8080 says:", "Initializing repository
installation failed", with an OK button. There is no error message visible
in paster.log. Here is what I do see:
127.0.0.1 - - [12/Sep/2014:11:45:57 -0700] "GET
ad0de2d&changeset_revisions=fe95cf1ae864 HTTP/1.1" 200 - "-"
ntosh; Intel Mac OS X 10_9_4) AppleWebKit/537.36 (KHTML, like Gecko)
DEBUG 2014-09-12 11:46:05,937 Creating repository dependency objects...
tool_shed.util.shed_util_common DEBUG 2014-09-12 11:46:07,214 Updating an
ing row for repository 'xena_import' in the tool_shed_repository table,
set to 'New'.
DEBUG 2014-09-12 11:46:07,241 Skipping installation of revision
repository 'start_xena' because it was installed with the (possibly
evision 75c7d80df9c1 and its current installation status is 'Installed'.
DEBUG 2014-09-12 11:46:07,241 Building repository dependency
127.0.0.1 - - [12/Sep/2014:11:46:05 -0700] "POST
stall HTTP/1.1" 200 - "
e2d&changeset_revisions=fe95cf1ae864" "Mozilla/5.0 (Macintosh; Intel Mac OS
0_9_4) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/37.0.2062.94
certificate with fingerprint
:e6:f8:5c:1c:b8:3e:8d:89:f0:0e:88:9a:ac:2b:61 not verified (check
nts or web.cacerts config setting)
127.0.0.1 - - [12/Sep/2014:11:46:07 -0700] "POST
ories HTTP/1.1" 500 - "
" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_9_4) AppleWebKit/537.36
ke Gecko) Chrome/37.0.2062.94 Safari/537.36"
Debug at: http://127.0.0.1:8080/_debug/view/1410547505
Then when I click OK, the status of the tool changes to "Cloning", and
stays in that state indefinitely while paster.log keeps logging a repeat of
127.0.0.1 - - [12/Sep/2014:11:55:28 -0700] "POST
tallation_status_updates HTTP/1.1" 200 - "
prepare_for_install" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_9_4)
t/537.36 (KHTML, like Gecko) Chrome/37.0.2062.94 Safari/537.36"
Just in case this is from some cruft from a previous installation, I tried
starting over with a brand new galaxy-dist directory. It doesn't look like
that changed anything.
Any ideas? Many thanks in advance!
On Fri, Sep 12, 2014 at 10:30 AM, Dave Bouvier <dave(a)bx.psu.edu> wrote:
The revision recorded in the filesystem path can seem odd at first glance.
My inline responses should hopefully clarify that.
On 09/10/2014 05:49 PM, Melissa Cline wrote:
> Hi Folks,
> Okay, I really need debugging ideas on this one.
> I have three repositories I'm developing under the test toolshed.
> They're named start_xena, xena_import and xena_find_datasets (they're
> all under visualization). start_xena contains a simple tool dependency
> to a package named installXena. xena_import and xena_find_datasets
> contain complex dependencies to installXena. I've installed these tools
> on two computers. On one, everything works great - amazingly well, in
> fact! Kudos, Development Team! On the other computer, the env.sh files
> associated with the complex dependencies keep referencing the wrong
> version of start_xena/installXena.
> Here are the gory details.
> The start_xena tool is installed in
> I'm a bit confused about the revision part of the path, because the
> latest revision is different (75c7d80df9c1), and I have uninstalled and
> re-installed the tool since its last update. Its package, installXena,
> is installed under tool_dependencies at
> That directory contains all the files I'd expect to be there, and its
> env.sh file contains all the environment variables I'd intended to set.
> So far, so good.
> xena_import is installed on my computer at
> Its tool_dependencies.xml reference installXena under start_xena as:
> <?xml version="1.0"?>
> <package name="installXena" version="1.0">
> <repository toolshed="http://testtoolshed.g2.bx.psu.edu"
> name="start_xena" owner="melissacline" changeset_revision="
> (Aside: is it necessary to have the changeset_revision in there? I'm a
> bit worried about the maintenance task of updating it for all of my
> dependent tools every time I update start_xena, but as I'll explain,
> it's not clear that this revision info is getting used).
It is not necessary to populate the changeset_revision or toolshed
attributes. If you omit them, they will be automatically filled in based on
the most recent matching repository.
> The tool dependency dir for xena_import is at
> env.sh reads:
> if [ -f
> ] ; then .
> ; fi
> Notice that the wrong revision number is in there. The 06 revision was
> an older one. Needless to say, the files and environment variables I
> need aren't in the 06 directory.
> Where does this incorrect revision number come from? I've deleted all
> the 06 directories, repeatedly. I've deleted and reinstalled the tools,
> and verified that after deletion, the directories in question were in
> fact gone. I've tried resetting the metadata for these tools under the
> Galaxy Admin menu. Now I need new ideas. As I mentioned, all of this
> works beautifully on a second computer. It's as if the first Galaxy
> installation, on the first computer, got stuck with some outdated
> information, and I haven't figured out how to clear it out.
The 0676 revision number is based on the revision that was first
installed, and should never change. When you update the repository, the
database will record the new revision in a separate field, but the original
revision id needs to be recorded in order to find the env.sh file on the
> Please keep all replies on the list by using "reply all"
> in your mail client. To manage your subscriptions to this
> and other Galaxy lists, please use the interface at:
> To search Galaxy mailing lists use the unified search at: